AQA A-Level Computer Science: Big Data — Practice Questions

These are original questions written for Marlbridge, for revision and practice on this content. They are not reproduced past-paper questions, and they do not replicate the exam’s exact structure, question count or mark tariffs — examination boards hold copyright in their own papers. Use these alongside the official past papers available free from your board.

Related: Big Data study guide | Big Data revision notes

These questions cover section 4.11.1 of the AQA A-Level Computer Science (7517) specification. Big Data is assessed on Paper 2. Questions 1-7 use this scenario.

Scenario. A city’s transport authority collects data from 40,000 road sensors (a reading every second), from bus GPS trackers, from traffic cameras, from passengers’ journey-planner apps and from social media posts that mention delays. It wants to predict congestion and reroute buses in real time.

---

Section A: describing Big Data

1. Explain how each of volume, velocity and variety applies to the transport authority’s data. [6]

2. The specification says the most difficult aspect of Big Data is its lack of structure. Give two reasons why lack of structure is a problem. [2]

3. Explain why a relational database is not appropriate for storing all of this data. [3]

4. Explain why machine learning techniques are needed to make use of the data. [2]

Section B: distributed processing

5. The data does not fit on a single server. (a) State what must happen to the processing as a result. [1] (b) Explain how each of these features of functional programming makes it easier to write correct code that can be distributed across servers: (i) immutable data structures (ii) statelessness (iii) higher-order functions. [6]

6. The authority wants the average speed recorded by each sensor over one day. Describe how map and fold (reduce) could be used across several servers to calculate this. [4]

Section C: modelling the data

7. (a) Explain what is meant by the fact-based model. [2] (b) Write three separate facts that could be stored about bus 214. [3] (c) Explain one advantage of a fact-based model over storing each bus as a single row that is overwritten when its details change. [2]

8. A music streaming service stores data about users, songs and artists. Users follow artists and play songs; each song is by one artist. (a) Draw a graph schema for this data, labelling the nodes and edges. [4] (b) Give one property each for a user node and a song node. [2] (c) Explain the difference between an edge and a property in a graph schema. [2]

---

Answers

1. Volume: readings from 40,000 sensors every second plus video and app data is far too much to fit on a single server [2]. Velocity: data arrives as a continuous stream and must be processed within milliseconds to seconds to reroute buses in real time [2]. Variety: the data comes in many forms – structured sensor readings and GPS coordinates, unstructured text from social media, and multimedia from cameras [2].

2. Analysing the data is made significantly more difficult [1]; it cannot be stored in the row-and-column format a relational database requires [1].

3. Relational databases require data to fit a fixed row-and-column format [1], but camera footage and social media text do not fit neatly into fields [1]. The volume is too large for one server, and relational databases do not scale well across multiple machines [1].

4. The data is too large and too varied for people, or fixed queries, to find patterns [1]; machine learning can discern patterns, such as the conditions that lead to congestion, and extract useful information automatically [1].

5. (a) It must be distributed across more than one machine [1]. (b) (i) Immutable data cannot be changed once created [1], so servers working on the same data at the same time cannot interfere with each other and no locking is needed [1]. (ii) Stateless functions have no side effects and always give the same output for the same input [1], so a task can run on any server in any order, and a failed task can be re-run with the same result [1]. (iii) Higher-order functions such as map, filter and fold take a function as an argument [1], so the same function can be sent to every server and applied to that server’s portion of the data in parallel [1].

6. The readings are split across the servers [1]. On each server, map is applied to every reading to turn it into a pair (sensor ID, (speed, 1)) [1]. fold combines pairs with the same sensor ID by adding the speeds and the counts, first on each server and then across servers [1]. Finally, for each sensor, total speed ÷ count gives the average [1].

7. (a) Data is stored as a collection of facts [1], each of which captures a single piece of information [1]. (b) Any three single-piece facts, e.g. “bus 214 has registration KX26 ABC”; “bus 214 was at stop 77 at 08:14:05”; “bus 214 is assigned to route 9” [3]. (c) New information is added as new facts rather than overwriting old data [1], so the history is kept (e.g. every past location of the bus) and nothing is lost if a value changes [1].

8. (a)

(User) --follows--> (Artist)
(User) --plays-->   (Song)
(Song) --by-->      (Artist)

Three node types: User, Song, Artist [1]; follows edge [1]; plays edge [1]; by edge with correct direction [1]. (b) User: e.g. username, date joined, country [1]. Song: e.g. title, duration, release year [1]. (c) An edge represents a relationship between two nodes (e.g. a user follows an artist) [1]; a property is a piece of information about a single node (e.g. the artist’s name) [1].

---

Where marks are usually lost

• Defining the three Vs in general terms without applying them to the scenario.
• Saying a relational database “cannot hold that much data” without mentioning its row-and-column format or poor scaling across machines.
• Naming functional programming features without explaining how each helps distributed processing.
• Writing a “fact” that contains several pieces of information at once.

Frequently asked questions

Is veracity one of the Vs in AQA 7517? No. The AQA 7517 specification describes Big Data in terms of volume, velocity and variety only. Answer with those three unless a question introduces another term.

Do I need to write map and fold code for Big Data questions? Not usually in this section. You need to explain how higher-order functions let the same processing run on many servers. Writing and evaluating map, filter and fold is tested in functional programming (4.12).